An acceptable use policy turns organizational AI principles into practical rules: which tools employees may use, what data they may enter, and which uses or product behaviors are prohibited or require approval. section 3.2.1
Make the rules concrete for both employees and product teams:
- For employees, prohibit entering customer contracts into public chatbots; specify approved tools and data limits. section 3.2.1
- For products, define permitted behavior and approval gates. MediAssist could explain uncertainty and route patients to clinicians, but could not diagnose, recommend treatment, or dismiss possible emergencies. section 3.2.2
- Require Steering Committee and clinician decisions before starting patient-facing health features; require independent validation, guideline checks, and human access for proposed medication support. section 3.2.2
Review the policy at least annually, and connect it to existing privacy, security, and business-risk processes rather than creating a separate governance track. section 3.2.1


